dplyr: select, filter, mutate

This is an R Markdown document. Follow the link to learn more about R Markdown and the notebook format used during the workshop.

Setup

library(tidyverse)

## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.2 ──
## ✔ ggplot2 3.3.6      ✔ purrr   0.3.4 
## ✔ tibble  3.1.8      ✔ dplyr   1.0.10
## ✔ tidyr   1.2.0      ✔ stringr 1.4.1 
## ✔ readr   2.1.2      ✔ forcats 0.5.1 
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()

The data is from the Stanford Open Policing Project and includes vehicle stops by the Evanston police in 2017. We’re reading the data in from a URL directly.

police <- read_csv("https://raw.githubusercontent.com/nuitrcs/r-tidyverse/main/data/ev_police.csv",
                   col_types=c( "location"="c"))

dplyr

dplyr is at the core of the tidyverse. It is for working with data frames. It contains six main functions, each a verb, of actions you frequently take with a data frame. We’re covering 3 of those functions today (select, filter, mutate), and 3 more next session (group_by, summarize, arrange).

If you have a background in Excel, you probably will start thinking about equivalences in your mind. That’s great! If you have worked on base R (not tidyverse), don’t overthink it. What’s the most important thing to keep in mind right now? Each of these functions takes a data frame as the first input. Within the function call, we can refer to the column names without quotes and without $ notation.

We will begin with the select for choosing columns, and filter for choosing rows. Remember select is for columns, while filter is for rows. Let’s start!

Select: Choose Columns

The previous session covered the basics of select but there are many more options for how we can specify which columns to choose.

First, let’s remember what the column names are:

names(police)

##  [1] "raw_row_number"            "date"                     
##  [3] "time"                      "location"                 
##  [5] "beat"                      "subject_age"              
##  [7] "subject_race"              "subject_sex"              
##  [9] "department_id"             "department_name"          
## [11] "type"                      "violation"                
## [13] "citation_issued"           "warning_issued"           
## [15] "outcome"                   "contraband_found"         
## [17] "contraband_drugs"          "contraband_weapons"       
## [19] "search_conducted"          "search_person"            
## [21] "search_vehicle"            "search_basis"             
## [23] "reason_for_stop"           "vehicle_make"             
## [25] "vehicle_year"              "raw_DriverRace"           
## [27] "raw_ReasonForStop"         "raw_TypeOfMovingViolation"
## [29] "raw_ResultOfStop"

Recall, the select function takes as the first input a data frame, and then we can list one or more columns, names unquoted, that we want to select. The columns will be ordered in the order we specify them. Notice that the order in the output is different than the original order above.

select(police, outcome, date)

Ranges

There are a number of select helper functions and special syntax options that allow us to choose multiple columns.

First, we can use : for range, but with names in addition to numbers:

select(police, raw_DriverRace:raw_ResultOfStop)

The result is the same if we use numbers.

select(police, 26:29)

We can select the rightmost columns with last_col(). You can check about with names that it is the “last” column in our dataset.

select(police, last_col())

Last 4 columns (the input to the function is the offset # of columns from the right edge):

select(police, last_col(0:3))

## Warning in offset && n <= offset: 'length(x) = 4 > 1' in coercion to
## 'logical(1)'

You can think everything() as the opposite of selecting one or a few columns. But it is most useful when combined with other functions. After all, we use select because you don’t want the whole database but part of it! We will see the use of everything() and other selection operators below.

select(police, everything())

Excluding columns

We can also say which columns we don’t want by putting a - (minus) in front of the name. Think about it as subtracting or or more columns. The idea of subtracting makes a lot of sense – we have now 27 rather than the original 29 columns in the data.

select(police, -raw_row_number, -subject_age)

When using negated - column names, if we start with negations, it will get messy. Make sure to avoid the following code:

select(police, -raw_row_number, -subject_age, time:outcome) 

To both specify the columns wanted and exclude some that would otherwise be selected, the exclusions need to come at the end:

select(police, time:outcome, -subject_age) 

We don’t need to include -raw_row_number – it is not part of the range between time and outcome. If we do include it, we will still get the same result.

select(police, time:outcome, -subject_age, -raw_row_number) 

Reordering and renaming

We’ve already seen that columns will appear in the result in the order that we list the names.

Do you remember the everything() helper function? It can be useful if you want to pull a few columns over to the left so that they are the ones that show first when you look at the data:

select(police, outcome, everything())

Each column only gets included once, in the position that it first appears. So “outcome” becomes the leftmost column above and no longer appears in it’s original spot.

We can also rename columns while using select(). The syntax is new_name = old_name.

select(police, raw_id=raw_row_number, date, time)

or we can use rename() to only rename, without affecting which columns are included or their order (all of the columns are kept in the same order):

rename(police, raw_id=raw_row_number)

Remember, this doesn’t change police because we didn’t save the result. So far, we’ve just been printing the copy of the data frame that is returned by the function. If we want to change our data frame, we’d need to save the result back to the police object.We can also save the changes in another object with a different name such as police_new.

police_new <- rename(police, raw_id=raw_row_number)

We saved the same data with a different name for one column, while keep all other and their original order. But we can also save only a few variables. For convenience, we name police_short our new object.

police_short<- select(police, raw_row_number, date, time)

EXERCISE 3

Remember: run the cells above to load tidyverse and import the data.

Using select and/or rename as needed:

• Rename subject_age to age, subject_race to race, and subject_sex to sex, but keep the columns in their original order
• Exclude the department_id and department_name columns

Hint:

• Remember that you can chain dplyr commands together with %>%.*
• You don’t need to save the results. If you save the changes, use a different name such as police_something.

Matching names

We can also select by matching patterns in the names of the columns. The patterns to match are in quotes because they aren’t column names – just character data.

select(police, starts_with("contraband"))

select(police, ends_with("issued"))

select(police, contains("vehicle"))

We can also put a - in front of these helper functions to exclude columns:

select(police, -contains("subject"))

And there are even more select helper functions.

EXERCISE 4

Use select() to get a copy of police without the columns that start with “raw”.

Hint: If you mess up your police dataset, re-run the cell near the top of the file under the Data header and read the data in again fresh.

Selecting with Vectors or Functions

What if we have the names of the columns we want to select in a vector already? For example:

analysis_vars <- c("search_basis", "reason_for_stop")

Perhaps we built this vector programatically (we wrote code to determine the values, instead of typing them ourselves), so we can’t just rewrite it to:

select(police, search_basis, reason_for_stop)

If we just give the vector to select, it looks like we expect “analysis_vars” to be a column name in police. We get a warning:

select(police, analysis_vars)

This warning tells us what we should do instead, which is use all_of:

select(police, all_of(analysis_vars))

This makes it clearer that “analysis_vars” isn’t the name of a column in police.

What if we want to select columns of a certain type – for example, only the numeric columns?

select(police, where(is.numeric))

is.numeric is the name of a function. We just use the name without (). This function is applied to each column, and if it returns TRUE, then the column is selected. Like above with using a vector, we wrap the function we want to use in where to make it clear that we’re using a function, not looking for a column named “is.numeric”).

where can be used with any function that returns a single TRUE or FALSE value for each column.

Filter: Choose Rows

The filter() function lets us choose which rows of data to keep by writing expressions that return TRUE or FALSE for every row in the data frame. Recall from last session:

filter(police, date == "2017-01-02")

We can do complex conditions as we could do with []

filter(police, subject_race == "hispanic" & subject_sex == "female")

If we include multiple comma-separated conditions, they are joined with & (and). So this following is equivalent to the above.

filter(police, subject_race == "hispanic", subject_sex == "female")

EXERCISE 5

1. Filter police to choose the rows where location is 60201 or 60202
2. Filter police to choose the rows where location is 60201 or 60202, and subject_sex is “male”

Hints:

• The “or” operator is |; the “and” operator is &

Including Variable Transformations

When filtering, we can include transformations of variables in our expressions. To see this, we’ll use the built-in mtcars dataset.

Here’s what mtcars looks like:

mtcars

Now, let’s filter to see which cars have above average mpg:

filter(mtcars, mpg > mean(mpg))

Or which car has the most horsepower (hp):

filter(mtcars, hp == max(hp))

EXERCISE 6

Using mtcars, find the car with the minimum (min) displacement (disp) value:

Bonus: slice variants

Last session, we saw slice() briefly as a way to choose which rows we want by their integer index value. But, there are some useful variants on the slice function that help us select rows that have the maximum or minimum value of a particular variable:

slice_max(mtcars, hp)

By default it just gives us one row, but we can ask for more than one by setting the n argument:

slice_max(mtcars, hp, n=3)

We got 4 rows above because there was a tie at position 3. There’s an option with_ties that can change how ties are handled.

There’s also a minimum version:

slice_min(mtcars, disp)

Like we did with slice_max, can also specify the n argument.

Mutate: Change or Create Columns

mutate() is used to both change the values of an existing column and make a new column.

We name the column we’re mutating and set the value. If the name already exists, it will update the column. If the name doesn’t exist, it will create a new variable (column is appended at the end of existing columns).

mutate(police, vehicle_age = 2017 - vehicle_year) %>%
  select(starts_with("vehicle"))  # just to pick a few columns to look at

We can put multiple mutations in the same call to mutate, with the expressions separated by commas:

mutate(police, 
       vehicle_age = 2017 - vehicle_year,
       old_car = vehicle_year < 2000)

Within a call to mutate, we can refer to variables we made or changed earlier in the same call as well. Here, we create vehicle_age, and then use it to create vehicle_age_norm:

police %>% 
  mutate(vehicle_age = 2017 - vehicle_year, 
         vehicle_age_norm = ifelse(vehicle_age < 0,  # ifelse test condition
                                   0,  # value if true
                                   vehicle_age)  # value if false
         ) %>%  
  # below is just making it easier for us to see what we changed
  select(starts_with("vehicle")) %>%
  filter(vehicle_age < 0)

Side note: there is a tidyverse version of ifelse() called if_else() – with the underscore or low dash symbol – that works generally the same except it is stricter about checking data types.

mutate() can also change an existing column. The location column in the data contains zip codes, that were read in as numeric values. This means the leading zero on some zip codes has been lost. Convert location to character data, and add back in the leading 0 if it should be there.

Here I’ll change the location column twice in the same call with two different transformations:

police %>%
  mutate(location = as.character(location),  # first convert to character, then recode below
         location = ifelse(nchar(location) == 4,  # ifelse test (vector of TRUE and FALSE)
                           paste0("0", location), # value if TRUE
                           location)) %>%  # value if FALSE
  select(location) %>%  # selecting just the column we mutated to look at
  filter(startsWith(location, "0"))  # selecting a few rows to look at the change

Remember that when using mutate(), you’re operating on the entire column at once, so you can’t select just a subset of the vector as you would with []. This means more frequently using functions like ifelse() or helper functions such as na_if(), replace_na(), or recode().

na_if replaces an existing value with NA. replace_na does roughly the opposite: replaces NA with a new value.

mutate(police, vehicle_make = na_if(vehicle_make, "UNK"))

na_if() can only can check and replace one value at a time; it also can’t be used with any expressions (x <= 1) – only single values.

EXERCISE 7

If the column beat in police is “/” or “CHICAGO”, set it to NA instead using mutate().

Hint: it’s ok if you take two steps to do this.

Recap

You now can use select and filter to subset your data in a wide variety of ways, and mutate to update variables or create new ones.

Next session: the three other common dplyr “verb” functions for working with data frames: group_by, summarize, and arrange.